Game addiction

    Studying data from Steam

    by Sebastian Wolf

    Task

    Your client is a mental health expert from an NGO who is interested in understanding more about gaming and the potentially addictive effect it can have on some individuals. You are meeting the client in a few days and they would like you to extract and present insights from the Steam dataset to help them in their research.

    Please use whichever tools you feel the most comfortable with, but we do recommend Tableau which is a popular choice. Tableau is free for students and there is also a free trial available here.

    My approach in these slides

    • Explore the context
      • Geographic distribution of gamers
      • Evolution of gamers on Steam
      • Times when gamers are active
    • Analyze game addiction
      • Define game addiction in the dataset
      • Explore correlations of playing time with several engagement factors
      • Derive conclusions and suggest focus for further analysis

    Exploring the context

    • Geographic distribution of gamers
    • Evolution of gamers on steam
    • When gamers are active

    The Steam dataset includes gamers from the entire world

    • Yet, most steam users report their location to be in the United States and Russia
    • Explore user counts per country by hovering over the map.

    Steam gaming is social

    • The number of new friendships is proportional to the number of new users
    • We see 2-3 times as many new friendships as new users per day

    Gamers are more active on weekends and mornings

    • The figure shows the day of the week and the hour when gamers sign up for Steam
    • We see heightened activity from Saturday morning to Monday morning
    Total number of new accounts created by day of week and hour_created
    0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
    dayofweek_created
    1 20,712 24,111 26,832 31,342 34,095 36,218 38,641 40,540 42,115 43,042 43,233 42,606 39,839 35,783 31,119 26,809 22,926 20,648 19,307 17,880 16,454 14,946 14,116 13,566
    2 14,412 15,932 18,235 20,946 24,119 27,304 31,342 34,746 37,464 39,771 41,727 40,832 39,022 34,727 29,931 25,468 21,908 20,008 18,899 17,768 16,401 14,957 13,890 13,366
    3 14,313 16,409 18,892 21,698 24,832 28,003 32,019 34,880 37,742 39,369 41,241 41,093 38,825 34,088 29,350 25,058 21,991 20,085 19,911 18,395 17,032 15,379 14,445 14,331
    4 14,790 16,674 18,663 21,668 24,928 29,029 32,721 35,825 38,279 40,294 41,175 40,272 37,919 34,552 30,678 26,029 22,960 21,000 19,750 18,820 17,630 16,241 15,168 14,850
    5 15,722 17,339 19,468 22,754 25,457 28,868 32,343 35,709 38,432 41,012 42,140 42,238 40,476 36,321 32,086 27,216 23,704 21,386 20,126 18,942 17,688 16,515 15,676 15,540
    6 16,600 18,336 20,696 24,241 27,735 32,039 36,887 40,296 42,730 45,495 46,455 46,023 43,803 40,327 36,501 31,101 27,244 24,522 23,321 22,993 22,021 21,111 20,438 20,172
    7 21,960 24,851 29,457 33,788 37,520 40,902 43,537 46,180 48,280 50,307 51,070 48,617 45,448 41,444 36,958 31,822 27,636 24,587 22,707 22,155 21,191 20,189 19,576 19,505

    Analyzing game addiction

    • Define addiction for this analysis
    • Explore correlations with engagement factors in 5 hypothesis
    • Summarize findings
    • Derive conclusions and suggest focus for further analysis

    Game addiction affects only a small fraction of gamers

    • The median playtime among active gamers (users that were active in the 2 weeks prior to data retrieval) was only 30min per day
    • Yet, 25% of active gamers play nearly 2h per day
    • We will focus on the 1% of all gamers that play more than 5h per day on average
    • To learn about game addicition, we will explore correlations of playtime with several engagement factors

    Hypothesis 1: The more friends I have, the more I play

    • Heavy gamers have more friends than regular gamers: average = 40 vs 7 and median 20 vs 1
    • There is a positive, but weak correlation of 0.3 between playtime and the number of friends
    • However, among heavy gamers, the friend count does not correlate with playtime
    • Takeaway: Heavy gamers generally have many friends on the network. But the number of friends does not correlate with their behaviour at the margin.

    Hypothesis 2: The more my friends play...

    • Heavy gamers' friends (if active) play more than regular gamers' friends (~2h vs 1h per day)
    • The same holds for the median friend (1h45min vs 10min)
    • For heavy gamers, playtime correlation with friends is only 0.03, vs 0.24 for the regular gamer
    • Takeaway: Friends seem to be an influence, but less so at the margin.

    Hypothesis 3: The more my heaviest gamer friend plays...

    • Correlation of playtime with the playtime of the friend that plays the most (0.29) is stronger than with the playtime of the average friend (0.24)
    • But again, this correlation is weaker for the heavy gamers (0.05)
    • Takeaway: The distribution of behaviour among friends matters. The maximum appears more important than the average.

    Hypothesis 4: The more groups I am part of...

    • Rather than friends, it could be the diversity of groups that keeps users engaged
    • Yet, group membership is uncommon: The median gamer is part of 0 groups vs 1 for heavy gamers
    • Correlation with playtime is very low
    • Takeaway: Groups seem to be less important than friends in understanding game addiction

    Hypothesis 5: The more multiplayer games I play...

    • It would seem that multiplayer games are more addictive, especially given the correlation of playtime and friends
    • However, there is nearly no correlation between the fraction of multiplayer games played and playtime
    • Takeaway: Regular gamers alike heavy gamers play a mix of single and multiplayer games

    What we learn from correlations

    • Several factors have been found to correlate with playtime:
      • The number of friendships
      • The playtime of friends
      • The maximum playtime among friends
      • The number of group memberships
    • Correlation is not causation: We don't know whether the above factors cause addiction
    • But we do know that these factors form important parts of heavy gamer's gaming experiences and need to be explored to understand addiction
    • We also found that these factors seem to play a lesser role in explaining why heavy gamers play an hour more or less (low correlations at the margin)

    Correlations between playtime and engagement factors are lower for heavy gamers

    Conclusions

    • Game addiction is rare among steam users (playtime distribution follows a power law)
    • The social factor (friends, groups) seems to be key to increased user engagement
    • To study addiction further, we need more information regarding gamers' interaction with games